They are running the spatial controls with that measurement, or otherwise their analysis is bunk: "The response then is generally within about +/-5db and varies little--after compensating the mic position a little - with any setting of the spatial control." If you look at the spatial control EQ in Graph 2, the response is +/- 10db. Therefore it has to be turned on when they are running that test for Graph 5, or there can be no +/-5db response once it's added in. Then the "setting of the spatial control" they are talking about refers to the low filter and bandwidth effects depicted in Graph 1 and maybe the loudness contour in Graph 3. Based on applying those graphs and graph 5, the conclusion of the +/-5db overall response makes sense.